Quantifying the Climate Impacts of Albedo Changes Due to Biofuel Production: a Comparison with Biogeochemical Effects

Quantifying the climate impacts of albedo changes due to biofuel production: a comparison with biogeochemical effects

The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters.

Citation Caiazzo, Fabio, Robert Malina, Mark D Staples, Philip J Wolfe, Steve H L Yim, and Steven R H Barrett. “Quantifying the Climate Impacts of Albedo Changes Due to Biofuel Production: a Comparison with Biogeochemical Effects.” Environmental Research Letters 9, no. 2 (January 1, 2014): 024015.

As Published http://dx.doi.org/10.1088/1748-9326/9/2/024015

Publisher Institute of Physics Publishing

Version Final published version

Citable link http://hdl.handle.net/1721.1/86306

Detailed Terms http://creativecommons.org/licenses/by/3.0/ Home Search Collections Journals About Contact us My IOPscience

Quantifying the climate impacts of albedo changes due to biofuel production: a comparison with biogeochemical effects

This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Environ. Res. Lett. 9 024015 (http://iopscience.iop.org/1748-9326/9/2/024015)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 18.51.1.88 This content was downloaded on 03/04/2014 at 15:31

Please note that terms and conditions apply. Environmental Research Letters Environ. Res. Lett. 9 (2014) 024015 (10pp) doi:10.1088/1748-9326/9/2/024015 Quantifying the climate impacts of albedo changes due to biofuel production: a comparison with biogeochemical effects

Fabio Caiazzo, Robert Malina, Mark D Staples, Philip J Wolfe, Steve H L Yim and Steven R H Barrett

Laboratory for Aviation and the Environment, Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cambridge, MA, USA

E-mail: [email protected]

Received 17 October 2013, revised 16 January 2014 Accepted for publication 17 January 2014 Published 26 February 2014

Abstract Lifecycle analysis is a tool widely used to evaluate the climate impact of greenhouse gas emissions attributable to the production and use of biofuels. In this paper we employ an augmented lifecycle framework that includes climate impacts from changes in surface albedo due to land use change. We consider eleven land-use change scenarios for the cultivation of biomass for middle distillate fuel production, and compare our results to previous estimates of lifecycle greenhouse gas emissions for the same set of land-use change scenarios in terms of CO2e per unit of fuel energy. We ﬁnd that two of the land-use change scenarios considered demonstrate a warming effect due to changes in surface albedo, compared to conventional fuel, the largest of which is for replacement of desert land with salicornia cultivation. This corresponds to 222 gCO2e/MJ, equivalent to 3890% and 247% of the lifecycle GHG emissions of fuels derived from salicornia and crude oil, respectively. Nine of the land-use change scenarios considered demonstrate a cooling effect, the largest of which is for the replacement of tropical rainforests with soybean cultivation. This corresponds to −161 gCO2e/MJ, or −28% and −178% of the lifecycle greenhouse gas emissions of fuels derived from soybean and crude oil, respectively. These results indicate that changes in surface albedo have the potential to dominate the climate impact of biofuels, and we conclude that accounting for changes in surface albedo is necessary for a complete assessment of the aggregate climate impacts of biofuel production and use.

Keywords: albedo, biofuels, climate, lifecycle S Online supplementary data available from stacks.iop.org/ERL/9/024015/mmedia

1. Introduction US EPA 2013). In the US, biofuel production for transportation aims to replace 30% of petroleum consumption by 2030 Biofuels may hold promise to promote energy security, re- (Perlack et al 2005, US Department of Energy 2011). Targets duce the environmental impact of transportation and foster are also set for the EU (10% replacement of diesel and gasoline economic development. For these reasons, many countries by 2020; EU 2009) and other countries such as China (2 million have enacted policies to encourage their production (EU 2009, tons of biodiesel by 2020; Koizumi 2011) and Indonesia (20% replacement of diesel and gasoline by 2025; Zhou and Content from this work may be used under the terms of Thomson 2009). the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the Historically, environmental assessments of biofuels have title of the work, journal citation and DOI. focused on biogeochemical effects (i.e. greenhouse gas (GHG)

1748-9326/14/024015+10$33.00 1 c 2014 IOP Publishing Ltd Printed in the UK Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al emissions) directly or indirectly attributable to the lifecycle quantify and compare the albedo effect of replacing an original of the fuel. Emissions are considered for all relevant life- land use with the cultivation of palm, rapeseed or salicornia, cycle steps, including feedstock cultivation, extraction, and in particular. The LUC scenarios considered are derived from transportation as well as fuel production, distribution, and Stratton et al (2010, 2011a) and the albedo effects are presented combustion (Kim and Dale 2005, Larson 2006, Lardon et al in terms of gCO2e/MJ of renewable middle distillate (MD) 2009, Yee et al 2009, Stratton et al 2010, Van der Voet fuel, which is the fuel considered in the Stratton et al (2011a) et al 2010, Guinee´ et al 2011). Land-use change (LUC) to lifecycle analysis. This enables a consistent comparison of cultivate biomass feedstock for biofuel production may lead the LUC-induced albedo effect, the biogeochemical effects to GHG emissions if it changes the amount of carbon stored from LUC, and the GHG emissions from the production of in vegetation and soil (Stratton et al 2010). renewable MD fuels. Distinct from assessing the biogeochemical effects, there is limited research focused on the biogeophysical effects 2. Methodology of LUC for biomass feedstock cultivation. Biogeophysical effects include changes in surface albedo (Betts 2000, Lee In this study, we evaluate the induced albedo effect of a number et al 2011), evapotranspiration (Pitman et al 2009, Georgescu of discrete LUC scenarios. Each of these scenarios is evaluated et al 2011), surface roughness/canopy resistance (Lean and at multiple geographic locations in order to account for Rowntree 1993, Betts 2007, Georgescu et al 2009), leaf area variability in surface and meteorological conditions within the index and rooting depth of the vegetation (Georgescu et al same land types involved in the LUC. Satellite measurements 2009). Of these, the LUC-induced change in surface albedo of albedo and transmittance parameters are retrieved for each geographic location of interest, and an analytical radiative is considered the dominant biogeophysical effect at the global balance model is used to convert the albedo changes into a scale (Betts 2000, 2001, Claussen et al 2001, Bala et al 2007). RF, then into equivalent GHG emissions. A change in albedo alters the surface reflectivity of sunlight (the incoming shortwave radiation), thus changing the Earth’s radiative balance. Albedo changes can be quantified in terms 2.1. LUC scenarios of global radiative forcing (RF) (Betts 2000, Georgescu et al We consider eleven LUC scenarios, comprised of five different 2011, Bright et al 2012, Cherubini et al 2012), which can be biomass feedstocks for up to three original land uses each, as expressed in terms of GHG equivalent emissions (Betts 2001, shown in table1. Each scenario is restricted to one geographic Bird et al 2008). This allows for a direct comparison against region. The scenarios are consistent between this study and the the biogeochemical effects calculated by traditional LCA. traditional LCA study that we use as a reference (Stratton et al In contrast, additional biogeophysical effects such as 2010; LUC combinations S1, S2, P1, P2, P3, H1), wherever evapotranspiration and surface roughness cannot be ade- possible. The switchgrass and rapeseed scenarios are redefined quately expressed in terms of global RF (Davin et al 2007, due to ambiguities in the reference LCA study. In Stratton et al Betts 2011, Cherubini et al 2012), although they may be (2010), switchgrass cultivation is assumed to take place on relevant at a local scale (Bounoua et al 2002, Georgescu et al generic carbon-depleted soils. In this study we consider three 2011). In previous work, the climate impact of albedo changes possible LUC scenarios associated with carbon-depleted soils has been assessed to describe the effect of forestation policies (McLaughlin et al 2002, Adler et al 2007): corn cultivation (Rautiainen et al 2009, Lohila et al 2010, Rautiainen et al (LUC B1), soybean cultivation (LUC B2), and barren land 2011). Recent studies have also attempted to evaluate the (LUC B3) replaced by switchgrass cultivation. Furthermore, in albedo effect of biomass feedstock cultivation, using either the reference LCA rapeseed cultivation in Europe is assumed numerical models (Georgescu et al 2011, Anderson-Teixeira to take place on set-aside land, i.e. land areas temporarily et al 2012, Hallgren et al 2013, Anderson et al 2013) or satellite removed from agricultural production (Stratton et al 2010). In measurements (Bright et al 2011, Loarie et al 2011, Cherubini this case we consider two LUC scenarios: corn cultivation (LUC R1) and uncultivated land (LUC R2) replaced by et al 2012). The results of those analyses suggest that albedo rapeseed cultivation. Table1 also indicates the geographic effects are potentially as important as the biogeochemical region in which each LUC scenario is assumed to take place effects assessed by traditional LCA (Georgescu et al 2011, in the reference LCA (Stratton et al 2011a). et al Anderson-Teixeira 2012). The assessments that are A minimum of four geographic locations are selected available in the literature often focus only on a single feedstock to describe each original land use type, and a minimum of (Georgescu et al 2011, Bright et al 2011, Loarie et al 2011) eight combinations of biomass feedstock and original land use and are based on different methodologies, making cross-study locations are used to define each of the 11 LUC scenarios comparison difficult. shown in table1. This multi-location approach allows for a In this study we perform an assessment of the LUC- more complete picture of the potential land conversions and induced albedo effects of a range of LUC scenarios by the associated natural variability. The latitude and longitude considering the cultivation of five different biomass feedstocks of these locations are retrieved using current literature (e.g., (switchgrass, soybean, palm, rapeseed and salicornia) and Mosali et al 2013), satellite observations (e.g., Rhines 2008) compare these effects to the biogeochemical effects quantified and farming databases (e.g., FIC 2013), and are confirmed by traditional LCA. To the best of our knowledge, this is the using the Moderate Resolution Imaging Spectroradiometer first study to consider the albedo effects from a broad range of (MODIS) Land Cover Type database (MCD12Q1) (NASA feedstocks using direct satellite measurements, and the first to MODIS 2013a).

2 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al

Table 1. Biomass feedstock types (ﬁrst column) and original land uses from this study (second column), compared to the reference LCA from Stratton et al (2010) (fourth column). Each LUC scenario for which the albedo effect is calculated is associated to a LUC code (third column). Geographic regions (ﬁfth column) are consistent with the reference LCA (Stratton et al 2010) for each LUC scenario, in order to enable comparison between albedo and biogeochemical effects.

Biomass Original land use LUC Original land use feedstock type (this study) code (reference LCA)a Region Switchgrass Corn cultivation B1 Central US Soybean cultivation B2 Carbon-depleted soil (Midwest-Northeast Barren land B3 states) Soy Cerrado grassland S1 Cerrado grassland Brazil (Central and Tropical rainforest S2 Tropical rainforest Southern regions) Palm Previously logged-over forest P1 Previously logged-over forest Southeast Asia Tropical rainforest P2 Tropical rainforest (Malaysia and Peat land rainforest P3 Peat land rainforest Indonesia) Rapeseed Corn cultivation R1 Europe (United Kingdom, Set-aside land Uncultivated land R2 France and Denmark) Salicornia Desert H1 Desert Mexico (Sonora desert) and US (Southern states)

a Stratton et al (2010).

2.2. Albedo and transmittance data retrieval and calculation i, the planetary albedo change is computed as a function of the day of the year d: Black-sky shortwave (BSW) broadband albedo coefficients (encompassing both the near infrared and visible spectra) are 1αi (d) = KT orig,i (d)Ta[αbio,i (d) − αorig,i (d)] (1) retrieved for each biomass feedstock and original land use pairing, i, that represents a specific LUC scenario. BSW albedo where αbio,i and αorig,i are the daily cloud-free shortwave coefficients are obtained from the MODIS satellite database albedo coefficients for biomass feedstock cultivation and the MCD43A3 (NASA MODIS 2013b) and vetted using a separate original land use, respectively, obtained from the MODIS MODIS database (MCD43A2; NASA MODIS 2013c) Albedo database (NASA MODIS 2013b) and averaged in time and data from the MODIS database are produced every 8 days and space as previously described. KT orig,i is the mean daily are linearly interpolated to obtain daily albedo evaluations for all-sky clearness index for the original land use point, and a full year. In case of missing or low quality data, the time Ta is the transmittance factor, as in Bright et al (2012). The interpolation is performed between the two closest acceptable calculation of planetary albedo differences from clear-sky observations. The average daily BSW albedo for each i is surface albedo values in equation (1) follows a procedure obtained for a full year by averaging the daily values retrieved widely reported in the literature (Lenton and Vaughan 2009, for three reference years (2009, 2010 and 2011), in order to Munoz˜ et al 2010, Bright et al 2012, Cherubini et al 2012). K account for annual variability in local conditions. Biases in Local mean daily values of T orig,i are constant for each month and are retrieved from the NASA Atmospheric Science each individual albedo observation are reduced by taking the and Data Center (ASDC) database, which provides monthly space average across the 500 m × 500 m cell where the location 22-year averages of the all-sky clearness index (including under investigation is found and the eight cells surrounding maximum and minimum bounds) (NASA ASDC 2013). The it. Consistency between the land-use type of these cells is transmittance factor T is chosen as a global annual average verified using the MODIS Land Cover Type database (NASA a of 0.854, consistent with previous findings and modeling MODIS 2013a). Table2 shows the yearly-averaged BSW comparisons (Lenton and Vaughan 2009, Munoz˜ et al 2010, albedo, retrieved and processed as described, for each land use Cherubini et al 2012, Bright and Kvalevag˚ 2013). type considered in the LUC scenarios from table1. The BSW albedos in table2 are given as mean, minimum and maximum values among the yearly-averaged albedos retrieved for all the 2.3. Radiative forcing (RF) model sample locations representing each specific land use type. The planetary albedo change due to biomass feedstock culti- For each biomass feedstock type and original land use vation on land originally used for some other purpose, found pairing i, representing a LUC scenario from table1 we evaluate in (1), alters the radiative balance of the Earth which can the albedo effect as the difference in RF induced by the be quantified as a radiative forcing. This is equal to the conversion of the original land use to biomass feedstock time integral of the product of daily albedo variation (1) and cultivation. The geographical location and conditions of ra- daily radiative flux at the top of the atmosphere RTOA,i (d), diative transmittance are kept the same as that of the original calculated at the original land use location (Bright et al 2012, land use; i.e. feedbacks between albedo changes and local Cherubini et al 2012). For each biomass feedstock and original weather/cloudiness conditions are not accounted for. For each land use pairing,i, the yearly global RF (measured in W m−2)

3 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al is therefore: clearness index from the ASDC database (NASA ASDC ( 365 ) 2013). Variability in meteorological conditions is therefore 1 X Aa accounted for. 1RF i = − [R i (d) · 1αi (d)] · (2) global, 365 TOA, A The geographic locations and relevant physical param- = earth d 1 eters for all sample locations are given in the Supporting where Aa is the reference area subject to the albedo change, and Information (SI available atstacks.iop.org/ERL/9/024015/m Aearth is the total area of the earth. The RF associated with each media). A more detailed derivation of the albedo-emission of the LUC scenarios in table1, 1RFLUC, is calculated as the conversion model and the yields and energy efﬁciencies of each average of all the global yearly radiative forcings 1RFglobal,i biomass feedstock type are also discussed in the SI (available found for all of the pairings, i representing the same LUC case. atstacks.iop.org/ERL/9/024015/mmedia).

2.4. CO2-equivalent emission conversion 3. Results and discussion

CO2e emissions per unit energy of biofuel produced (gCO2e/ The albedo effect for the eleven LUC scenarios is shown in MJ) is a common metric adopted in LCA studies (Larson figure1 (blue bars) in terms of emissions or sequestrations 2006, Adler et al 2007, Stratton et al 2010, 2011a). To of CO2e per unit energy of biofuel produced (gCO2e/MJ). establish direct comparison between albedo change effects The whisker bars represent low and high cases, corresponding and biogeophysical effects for the same LUC scenario, the to the minimum and maximum RF from the albedo changes global RF associated with each one of the LUC scenarios in induced by each LUC scenario, taking into account the variability of the albedo and meteorological conditions among table1 is converted into CO 2e. The correspondence between the locations representative of the same LUC scenario. the RF induced by albedo changes and CO2e emissions is well established in the literature (Betts 2000, Bird et al 2008, The red bars in figure1 show the biogeochemical impacts Munoz˜ et al 2010, Joos et al 2013). First, RF is converted into calculated in the reference LCA by Stratton et al (2010) for the same biomass feedstock cultivation and original land use pair- a change in atmospheric carbon concentration 1C by using a ings. The biogeochemical effects calculated in the reference logarithmic relation with the background carbon concentration study include GHG emissions from cultivation, harvesting, (Betts 2000) (linearized for small perturbations). Positive RF extraction and transportation of the biomass; processing of (induced by a decrease in the land albedo) corresponds to biomass into MD fuels; and transportation, distribution, and carbon emissions, while negative RF corresponds to carbon combustion of the finished fuel product. Non-CO2 GHGs sequestrations. The concentration 1C (in parts per million, and emissions species from direct fuel combustion are not ppm) is then converted into an equivalent carbon emission considered in the reference LCA. In the case of aviation, 1CT per unit area subject to albedo change: these effects can result in a doubling of CO2e direct fuel combustion emissions for a 100 year time horizon (Stratton 1 MC 1CT = 1Cmatm et al 2011b). Only GHG emissions associated with direct AFTH=100 Mair LUC are considered in the reference LCA (Stratton et al 1 1 × −1 2010). Emissions from indirect LUC, which occurs if direct 6 (tC ha ) (3) Aa 10 LUC disrupts the equilibrium between supply and demand for the displaced crop, and for downstream products relying where matm is the total mass of the atmosphere (in tons), on this crop (Plevin et al 2010), are not taken into account. MC and Mair are the molecular weights of carbon and air The low and high ranges for the red bars in figure1 reflect −1 respectively (in g mol ), Aa is expressed in hectares and the variability of parameters used for LCA, such as process AFTH=100 is the airborne emission fraction of CO2 for a efficiency and biomass feedstock yield (Stratton et al 2010). time horizon (TH) of 100 years, consistent with the reference The green bars in figure1 represent the sum of albedo and biogeochemical impacts assessment by Stratton et al (2010). biogeochemical effects. This net effect can be compared to Finally, using data about the biomass yield, mass conversion the reference lifecycle emissions for conventional MD (90 factor, and specific energy conversion efficiencies, the carbon gCO2e/MJ, dashed black line in figure1), assumed to be emission per unit area in (3) is converted into a CO2e equal to the results for conventional diesel from Stratton et al emission per unit energy of the fuel produced. In order (2011a). We do not consider albedo effects attributable to to evaluate the albedo effects under the same assumptions conventional middle distillate fuels, since in this case land as the biogeochemical impacts, the resulting emissions are use change per unit energy of finished fuel is estimated to be distributed over 30 years, as in the reference LCA (Stratton two to three orders of magnitude lower than for biomass-based et al 2010). fuels (Yeh et al 2010). In order to represent the magnitude of natural variability, we examine low, baseline, and high cases for each LUC 3.1. Switchgrass scenario in table1. The baseline cases utilize the mean of the The albedo change due to replacement of corn cultivation, RFs calculated for all the biomass feedstock and original land soy cultivation and barren land with switchgrass (scenarios use pairings representing the same LUC scenario. Low and B1–B3) leads to a negative RF in the baseline results, high cases utilize the maximum and minimum RFs calculated the equivalent of a sequestration of CO2e. The effect is among the pairings used to simulate the same LUC scenario, stronger when switchgrass replaces corn (LUC B1, −22 and the upper and lower estimates of the relevant all-sky gCO2e/MJ) or soy (LUC B2, −13 gCO2e/MJ) than it is for

4 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al

Figure 1. Climate impacts of biofuel production and use for different LUC scenarios. Each row of the table on the left contains a biomass feedstock type and original land use pairing, corresponding to a particular MD fuel, indicated in the last column. In the histogram, the blue bars indicate the impact of the albedo variations due to each LUC, in terms of CO2e per unit energy of fuel produced. The high and low cases include variability in the geographical locations and local meteorological conditions (as described in section 2.4). The red bars show the biogeochemical effects (i.e., the direct GHG emissions) as calculated in the reference LCA (Stratton et al 2010). The related whisker bars account for the variability of process efﬁciency and biomass feedstock yield (Stratton et al 2010). The green bars in the background show the net impact of considering albedo and biogeochemical effects in the baseline, low and high emission scenarios. Both albedo and biogeochemical effects are distributed over a time span of 30 years, consistent with Stratton et al (2010). The dashed black line indicates the results for conventional MD fuel from the reference LCA (Stratton et al 2010).

barren land (LUC B3, −9 gCO2e/MJ). A negative RF (or soybean cultivation results, in both cases, in a negative RF, equivalently a cooling effect) for the conversion of land to equivalent to a cooling effect (−146 and −161 gCO2e/MJ, switchgrass cultivation has also been computed by Georgescu respectively). Soybean cultivation in Brazil generally exhibits et al (2011)(−0.0053 W m−2) and Anderson-Teixeira et al a higher reflectivity (average albedo of 0.175, see table2) than −1 −1 (2012)(−50 Mg CO2e ha 50 yr ) however the disparate the savannah land of the Brazilian Cerrado (average albedo methodologies, metrics and LUC assumptions used in those of 0.133) or than the dense dark tropical rainforest in the studies do not allow a quantitative comparison with this Amazon region (average albedo of 0.120). An albedo-induced work. The biogeochemical lifecycle impacts of renewable CO2e reduction due to the establishment of soy cultivation in MD fuel production and use from switchgrass replacing Brazil has also been found by Anderson-Teixeira et al (2012) −1 −1 carbon-depleted soils are estimated by Stratton et al (2010) (−70 MgCO2e ha 50 yr ). The green bar in figure1 for to be nearly carbon neutral (−1.6 gCO2e/MJ). In absolute LUC S1 shows that, considering both the effects of albedo terms, the albedo impact of switchgrass cultivation for MD change and of biogeochemical lifecycle GHG emissions cal- production is therefore 1380%, 813% and 563% greater than culated by the reference LCA (Stratton et al 2010), the aggre- the biogeochemical lifecycle effects of corn, soybean, and gate climate impact of the conversion of cerrado grassland barren land conversion, respectively. The net sum of albedo to soybean cultivation is equivalent to a sequestration of and biogeochemical effects is −23 gCO2e/MJ, −15gCO2e/MJ 50 gCO2e/MJ. By adding the albedo effect to the biogeo- and −11 gCO2e/MJ for corn replacement (LUC B1), soy chemical results from the reference LCA, renewable MD from replacement (LUC B2) and barren land replacement (LUC soybean cultivated on land that was previously Cerrado grass- B3), respectively. It should be noted that the high emission land exhibits a decrease in climate impact of 156% with respect cases (high whisker bar limits) demonstrate that a net positive to conventional MD (from 90 to −50 gCO2e/MJ). This reverts RF is also possible for these three LUC scenarios. the results of the reference LCA which suggests that soybean derived MD has greater climate impact, in terms of lifecycle GHG emissions, than conventional MD (97 gCO e/MJ versus 3.2. Soy 2 90 gCO2e/MJ) (Stratton et al 2010). In the case of tropical In figure1, LUC S1 and S2 show that the albedo effect rainforest replacement with soybean cultivation (LUC S2), of replacing Brazilian Cerrado and tropical rainforest with inclusion of the albedo effects does not revert the findings of

5 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al

Table 2. Black-sky shortwave albedo for each land use type considered in this study. Albedo values are given as mean, minimum and maximum values among yearly-averaged BSW albedo coefﬁcients retrieved for the sample locations representing a speciﬁc land use type. The number of sample locations for each land use type is indicated in the Table. Yearly-averaged BSW albedos at each location are obtained from the MODIS satellite database MCD43A3 (NASA MODIS 2013b), and treated as described in section 2.2.

BSW albedo Land type Number of samples Mean Min Max Switchgrass field 4 0.177 0.156 0.191 (US) Soybean cultivation 6 0.175 0.139 0.204 (Brazil) Palm plantation 14 0.088 0.045 0.134 (SE Asia) Rapeseed field 11 0.162 0.147 0.179 (Europe) Salicornia cultivation 5 0.149 0.140 0.156 (Mexico) Corn field 5 0.165 0.151 0.178 (US) Soybean cultivation 8 0.163 0.148 0.200 (US) Barren land 4 0.175 0.135 0.222 (US) Cerrado grassland 4 0.133 0.122 0.148 (Brazil) Tropical rainforest 4 0.120 0.111 0.128 (Brazil) Previously logged-over forest 9 0.125 0.093 0.138 (SE Asia) Tropical rainforest 4 0.066 0.033 0.103 (SE Asia) Peat land rainforest 4 0.091 0.046 0.129 (SE Asia) Corn field 11 0.156 0.145 0.174 (Europe) Uncultivated land 8 0.151 0.132 0.184 (Europe) Desert land 10 0.326 0.251 0.535 (South US/Mexico)

the reference LCA: the direct GHG emission calculated by the of tropical rainforest (LUC S2, −161 gCO2e/MJ). The low reference LCA (569 gCO2e/MJ) (Stratton et al 2010) is 453% albedo of palm also accounts for the positive RF induced by larger than the impact attributable to the increase in albedo conversion of previously logged-over forest (LUC P1), equiv- (−161 gCO2e/MJ). alent to a GHG emission of 14 gCO2e/MJ. Palm replacement of peat land rainforest (LUC P3) yields a relatively small − 3.3. Palm sequestration of 4 gCO2e/MJ. For the aggregate climate impact of LUC P1, indicated by the green bar in figure1, the Due to their leaf characteristics and plantation density, palms baseline case of 55 gCO2/MJ remains below the conventional are characterized by the lowest average shortwave albedo jet fuel baseline even if the albedo effect is included. If the (0.088, see table2) among the biomass feedstock types consid- albedo effect is included, the high emission case (high whisker ered in this study. Furthermore, palm is not harvested, meaning bar limit, 129 gCO2e/MJ) is 43% larger than conventional that the low albedo coefficient is nearly constant throughout MD, whereas the high emission case for the biogeochemical the year. This accounts for the smaller cooling effect shown effects does not exceed the conventional MD reference. For in figure1 for the case of tropical rainforest replaced by LUC P2, while the baseline warming effect remains higher palm plantations (LUC P2, −25 gCO2e/MJ) compared to the than that of conventional MD even if the albedo effect is cooling effect occurring for soybean cultivation replacement included, the aggregate climate impact in the low emission

6 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al

Figure 2. Climate impacts of direct LUC for different LUC scenarios. Each row of the table on the left contains a biomass feedstock type and original land use pairing, corresponding to a particular MD fuel, indicated in the last column. In the histogram, the blue bars indicate the instantaneous effect of changes in albedo due to LUC, in terms of CO2 equivalent emissions per unit energy of fuel. The high and low cases include variability in the geographical locations and local meteorological conditions (as described in section 2.4). The red bars show the biogeochemical effects (i.e., LUC-induced GHG emissions) related exclusively to LUC, as calculated by Stratton et al (2010). The related whisker bars account for the variability in biomass feedstock yield (Stratton et al 2010). The green bars in the background show the aggregate climate impact of LUC considering albedo and biogeochemical effects in the baseline, low and high emission scenarios.

case (77 gCO2e/MJ) is below that of conventional MD, which by LUC H1, equivalent to 222 gCO2e/MJ. In comparison, the was not the case in the reference LCA. For LUC P3, the direct GHG emissions computed by the reference LCA are biogeochemical effects from the reference LCA are so large only 6 gCO2e/MJ (Stratton et al 2010). This result shows that (705 gCO2e/MJ) (Stratton et al 2010), that the inclusion production and use of renewable MD using salicornia grown on of albedo effects does not change the total climate impact desert lands can have a larger warming effect than producing significantly. and using conventional MD. Figure2 shows a comparison between albedo effects and 3.4. Rapeseed GHG emissions from LUC, excluding all the other stages accounted for in the reference LCA (Stratton et al 2010) (i.e. LUC R1 and R2 show the albedo effect of replacing corn biomass cultivation and transport; feedstock to fuel conver- or uncultivated land with rapeseed cultivation in Europe. In sion; and biofuel transport and combustion). This comparison both cases a small cooling effect is found, −3gCO e/MJ for 2 is instructive since it shows the relative magnitude of the bio- LUC R1 and −10 gCO e/MJ for LUC R2. According to the 2 geochemical and biogeophysical effects that exclusively stem reference LCA, the biogeochemical effects of renewable MD from (direct) alterations in land use. Both effects are evaluated fuel from rapeseed cultivated on previously set-aside land, as instantaneous CO e emissions, not distributed across the yields a biogeochemical effect of 96 gCO e/MJ (Stratton et al 2 2 reference 30-years time span as in figure1 (see equation S19 2010). The contribution of the albedo effect is negligible for in the SI available atstacks.iop.org/ERL/9/024015/mmedia) this set of LUC scenarios. but only across the first year of land use change. This is because the albedo effect is obtained using albedo differences 3.5. Salicornia and transmittance parameters averaged over a full year of LUC H1 considers salicornia cultivation on land that was variation, and carbon sequestration due to LUC is evaluated previously desert. The albedo of the desert (average of 0.326 after a full year of vegetation replacement (Stratton et al 2010). from table2) is much larger than the albedo of the salicornia Therefore, the results for albedo effects (blue bars) are directly cultivations (average of 0.149), since deserts are composed proportional to the ones shown in figure1. Results for LUC of smooth, clear sand, and are highly reflective of incoming emissions (red bars) are instead proportionally smaller than solar radiation (Pielke and Avissar 1990). The large desert the ones reported in figure1, since they exclude steady-state albedo accounts for the relatively large positive RF induced transport, production and combustion emissions. The results

7 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al indicate that albedo effects are on the same order of magnitude seasonal impacts on the local climate. These impacts may as the traditional direct biogeochemical LUC emissions for be offsetting when averaged over a whole year, as found most of the LUC scenarios considered. by Georgescu et al (2013) for the biogeophysical effects of savannah to sugarcane conversion. Finally, the albedo impacts 4. Conclusions of other variables such as snow cover variation due to LUC, This study shows that changes to surface albedo due to biomass and climate-meteorology feedbacks potentially affecting local cultivation can have a significant impact on the aggregate cloudiness, are not accounted for in this study. climate impact of biofuels. The albedo effects of LUC related to biomass feedstock cultivation for biofuel produc- Acknowledgments tion, shown in figure1, are on the same order of magnitude as the biogeochemical effects calculated by traditional LCA This work was made possible by funding from the Fed- for the same LUC scenarios. The largest effects are calcu- eral Aviation Administration (FAA), Air Force Research lated for LUC scenarios S1 and H1. Renewable MD pro- Laboratory (AFRL) and the Defense Logistics Agency-Energy duction from soybean cultivated on land that was previously (DLA Energy), under Project 47 of the Partnership for Air cerrado grassland (LUC S1) is found to yield a net cooling Transportation Noise and Emissions Reduction (PARTNER). effect, equivalent to −50 gCO2e/MJ. This makes renewable The authors would like to thank Dr James I Hileman and MD derived from soy oil a potentially viable alternative Dr Mohan Gupta at the FAA for their guidance on technical to conventional MD, a result that was not apparent when matters. Any views or opinions expressed in this work are the albedo effect was not included in the reference LCA those of the authors and not the FAA, AFRL or DLA-Energy. (Stratton et al 2010). Conversely, renewable MD production We also thank the reviewers for their comments. from salicornia cultivated on land that was previously desert yields a net warming effect, corresponding to 228 gCO2e/MJ References of MD fuel. This is the first evidence that salicornia-derived biofuel obtained by converting desert land could be potentially Adler P R, Grosso S J D and Parton W J 2007 Life-cycle assessment detrimental from a climate impact standpoint when compared of net greenhouse-gas flux for bioenergy cropping systems Ecol. to conventional fuels. Our results give support for further Appl. 17 675–91 evaluating the consideration of LUC-induced surface albedo Anderson C J, Anex R P, Arritt R W, Gelder B K, Khanal S, changes in global biofuels policies (Betts 2000, 2001, Claussen Herzmann D E and Gassman P W 2013 Regional climate impacts Geophys. Res. Lett. 40 et al 2001, Bala et al 2007). of a biofuels policy projection 1217–22 Anderson-Teixeira K J, Snyder P K, Twine T E, Cuadra S V, Some limitations of this study warrant acknowledgment. Costa M H and DeLucia E H 2012 Climate-regulation services of First, our analysis is restricted to changes in surface albedo, natural and agricultural ecoregions of the Americas Nature Clim. and other biogeophysical impacts such as evapotranspiration, Change 2 177–81 surface roughness and rooting depth are not quantified here. Bala G, Caldeira K, Wickett M, Phillips T J, Lobell D B, Delire C Second, the albedo effects shown in figure1 are dependent on and Mirin A 2007 Combined climate and carbon-cycle effects of the sample geographical locations chosen, and should not be large-scale deforestation Proc. Natl Acad. Sci. 104 6550–5 interpreted as characteristics of the feedstocks considered, but Betts R A 2000 Offset of the potential carbon sink from boreal rather as a function of the biomass feedstock and original forestation by decreases in surface albedo Nature 408 187–90 land use pairing investigated. Third, the use of equivalent Betts R A 2001 Biogeophysical impacts of land use on present-day emissions based on RF has a theoretical weakness (Davin climate: near-surface temperature change and radiative forcing et al 2007) because albedo effects and biogeochemical effects Atmos. Sci. Lett. 2 39–51 Betts R A 2007 Implications of land ecosystem-atmosphere act on different spatial and temporal scales. Nevertheless, interactions for strategies for climate change adaptation and equivalent emissions of CO2 per unit energy of combusted mitigation Tellus B 59 602–15 fuel is a widely accepted metric used to compare both effects Betts R A 2011 Mitigation: a sweetener for biofuels Nature Clim. (Betts 2001, Bird et al 2008, Georgescu et al 2011, Cherubini Change 1 99–101 et al 2012). When RF is used to compare different climate Betts R A, Falloon P D, Goldewijk K K and Ramankutty N 2007 change mechanisms, there is an implicit assumption that the Biogeophysical effects of land use on climate: model simulations climate response is proportional to forcing. However, research of radiative forcing and large-scale temperature change Agricult. by Hansen and Nazarenko (2004) show that surface forcing Forest Meteorol. 142 216–33 may be twice as effective at high latitudes as at low latitudes Bird D N, Kunda M, Mayer A, Schlamadinger B, Canella L and in generating surface temperature change. Further discussion Johnston M 2008 Incorporating changes in albedo in estimating is available in Betts et al (2007), Bird et al (2008), and the climate mitigation benefits of land use change projects Biogeosci. Discuss. 5 1511–43 in Cherubini et al (2013). With respect to time scales, the Bounoua L, DeFries R, Collatz G J, Sellers P and Khan H 2002 global RFs (and equivalent CO emissions) evaluated in this 2 Effects of land cover conversion on surface climate Clim. Change study reflect the impacts of albedo variations averaged over 52 29–64 a whole year of LUC in order to compare the albedo effect Bright R M, Cherubini F and Strømman A H 2012 Climate impacts with the long-term biogeochemical impacts. However, albedo of bioenergy: inclusion of carbon cycle and albedo dynamics in coefficients are dependent upon transient surface conditions life cycle impact assessment Environ. Impact Assess. Rev. (Song 1999), and these variations may lead to significant 37 2–11

8 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al Bright R M and Kvalevag˚ M M 2013 Technical note: evaluating a Lean J and Rowntree P R 1993 A GCM simulation of the impact of simple parameterization of radiative shortwave forcing from Amazonian deforestation on climate using an improved canopy surface albedo change Atmos. Chem. Phys. Discuss. 13 18951–67 representation Q. J. R. Meteorol. Soc. 119 509–30 Bright R M, Strømman A H and Peters G P 2011 Radiative forcing Lee X et al 2011 Observed increase in local cooling effect of impacts of boreal forest biofuels: a scenario study for norway in deforestation at higher latitudes Nature 479 384–7 light of albedo Environ. Sci. Technol. 45 7570–80 Lenton T M and Vaughan N E 2009 The radiative forcing potential Cherubini F, Bright R M and Strømman A H 2012 Site-specific of different climate geoengineering options Atmos. Chem. Phys. global warming potentials of biogenic CO2 for bioenergy: 9 5539–61 contributions from carbon fluxes and albedo dynamics Environ. Loarie S R, Lobell D B, Asner G P, Mu Q and Field C B 2011 Res. Lett. 7 045902 Direct impacts on local climate of sugar-cane expansion in Brazil Cherubini F, Bright R M and Strømman A H 2013 Global climate Nature Clim. Change 1 105–9 impacts of forest bioenergy: what, when and how to measure? Lohila A, Minkkinen K, Laine J, Savolainen I, Tuovinen J-P, Environ. Res. Lett 8 014049 Korhonen L, Laurila T, Tietav¨ ainen¨ H and Laaksonen A 2010 Claussen M, Brovkin V and Ganopolski A 2001 Biogeophysical Forestation of boreal peatlands: impacts of changing albedo and versus biogeochemical feedbacks of large-scale land cover greenhouse gas fluxes on radiative forcing J. Geophys. Res.: change Geophys. Res. Lett. 28 1011–4 Davin E L, De Noblet-Ducoudre´ N and Friedlingstein P 2007 Biogeosci. 115 G4 Impact of land cover change on surface climate: Relevance of the McLaughlin S B, De la Torre Ugarte D G, Garten C T Jr, Lynd L R, radiative forcing concept Geophys. Res. Lett. 34 L13702 Sanderson M A, Tolbert V R and Wolf D D 2002 High-value EU 2009 Renewable Energy Directive (Dir 2009/28/EC)http://eur-l renewable energy from prairie grasses Environ. Sci. Technol. ex.europa.eu/LexUriServ/LexUriServ.do?uri=Oj:L:2009:140:001 36 2122–9 6:0062:en:PDF Mosali J, Biermacher J T, Cook B and Blanton J 2013 Bioenergy FIC 2013 Farmland Information Center,www.farmlandinfo.org/, for cattle and cars: a switchgrass production system that engages accessed in May 2013 cattle producers Agron. J. 105 960 Georgescu M, Lobell D B and Field C B 2009 Potential impact of Munoz˜ I, Campra P and Fernandez-Alba´ A R 2010 Including US biofuels on regional climate Geophys. Res. Lett. 36 L21806 CO2-emission equivalence of changes in land surface albedo in Georgescu M, Lobell D B and Field C B 2011 Direct climate effects life cycle assessment. Methodology and case study on of perennial bioenergy crops in the United States Proc. Natl. greenhouse agriculture Int. J. Life Cycle Assess 15 672–81 Acad. Sci. 108 4307–12 NASA ASDC 2013 NASA Surface Meteorology and Solar Georgescu M, Lobell D B, Field C B and Mahalov A 2013 Energy—Location available athttps://eosweb.larc.nasa.gov/cgi-bi Simulated hydroclimaticimpacts of projected Brazilian sugarcane n/sse/grid.cgi? accessed in July 2013 expansion Geophys. Res. Lett. 40 972–7 NASA MODIS 2013a Index of /MOTA/MCD12Q1.051 available at Glenn E P, O’leary J W, Watson M C, Thompson T L and http://e4ftl01.cr.usgs.gov/MOTA/MCD12Q1.051/ accessed in Kuehl R O 1991 Salicornia bigelovii Torr.: an oilseed halophyte July 2013 for seawater irrigation Science 251 1065–7 NASA MODIS 2013b Index of /MOTA/ MCD43A3.005 available at Guinee´ J B, Heijungs R, Huppes G, Zamagni A, Masoni P, http://e4ftl01.cr.usgs.gov/MOTA/MCD43A3.005/ accessed in Buonamici R, Ekvall T and Rydberg T 2011 Life cycle July 2013 assessment: past, present, and future Environ. Sci. Technol. NASA MODIS 2013c Index of /MOTA/ MCD43A2.005 available at 45 90–6 http://e4ftl01.cr.usgs.gov/MOTA/MCD43A2.005/ accessed in Hallgren W, Schlosser C A, Monier E, Kicklighter D, Sokolov A July 2013 and Melillo J 2013 Climate impacts of a large-scale biofuels OECD/IEA 2004 World Energy Outlook 2004 expansion Geophys. Res. Lett. 40 1624–30 Hansen J and Nazarenko L 2004 Soot climate forcing via snow and Perlack R D, Wright L L, Turhollow A F, Graham R L, Stokes B J ice albedos Proc. Natl Acad. Sci. 101 423–8 and Erbach D C 2005 Biomass as feedstock for a bioenergy and Hill J, Nelson E, Tilman D, Polasky S and Tiffany D 2006 bioproducts industry: the technical feasibility of a billion-ton Environmental, economic, and energetic costs and benefits of annual supply US Department of Energy and US Department of biodiesel and ethanol biofuels Proc. Natl Acad. Sci. Agriculture Report ORNL/TM-2005/66 103 11206–10 Pielke R A and Avissar R 1990 Influence of landscape structure on Hungate B A and Hampton H M 2012 Ecosystem services: valuing local and regional climate Landscape Ecol. 4 133–55 ecosystems for climate Nature Clim. Change 2 151–2 Pitman A J et al 2009 Uncertainties in climate responses to past land Joos F et al 2013 Carbon dioxide and climate impulse response cover change: First results from the LUCID intercomparison functions for the computation of greenhouse gas metrics: a study Geophys. Res. Lett. 36 L14814 multi-model analysis Atmos. Chem. Phys. 13 2793–825 Plevin R J, Michael O’Hare, Jones A D, Torn M S and Gibbs H K Kim S and Dale B E 2005 Life cycle assessment of various cropping 2010 Greenhouse gas emissions from biofuels’ indirect land use systems utilized for producing biofuels: Bioethanol and biodiesel change are uncertain but may be much greater than previously Biomass Bioenergy 29 426–39 estimated Environ. Sci. Technol. 44 8015–21 Koizumi T 2011 Biofuel programs in East Asia: developments, Rautiainen M, Nilson T and Lukk¨ T 2009 Seasonal reflectance perspectives, and sustainability Environmental Impact of Biofuels trends of hemiboreal birch forests Remote Sens. Environ. ed M Aurelio Dos and S Bernardes (Rijeka: InTech Europe) 113 805–15 Lardon L, Helias´ A, Sialve B, Steyer J-P and Bernard O 2009 Rautiainen M, Stenberg P, Mottus M and Manninen T 2011 Life-cycle assessment of biodiesel production from microalgae Radiative transfer simulations link boreal forest structure and Environ. Sci. Technol. 43 6475–81 shortwave albedo Boreal Environ. Res. 16 91–100 Larson E D 2006 A review of life-cycle analysis studies on liquid Rhines S 2008 Oklahoma’s Bioenergy Future The Noble biofuel systems for the transport sector Energy Sust. Dev. Foundationhttp://www.ces.ncsu.edu/nreos/forest/feop/biomass-s 10 109–26 outh/proceedings/pdf/0923-1515c-Rhines.pdf

9 Environ. Res. Lett. 9 (2014) 024015 F Caiazzo et al Song J 1999 Phenological influences on the albedo of prairie Van der Voet E, Lifset R J and Luo L 2010 Life-cycle grassland and crop fields Int. J. Biometeorl. 42 153–7 assessment of biofuels, convergence and divergence Biofuels Stratton R W, Wong H M and Hileman J I2010 Lifecycle 1 435–49 PARTNER greenhouse gas emissions from alternative jet fuels Wicke B, Dornburg V, Faaij A P C and Junginger H M 2007 Project 28 Report, PARTNER-COE-2010-001 A Greenhouse Gas Balance of Electricity Production from Stratton R W, Wong H M and Hileman J I 2011a Quantifying variability in life cycle greenhouse gas inventories of alternative Co-Firing Palm Oil Products from Malaysia Universiteit Utrecht middle distillate transportation fuels Environ. Sci. Technol. Copernicus Institute NWS-E-2007–33 45 4637–44 Yee K F, Tan K T, Abdullah A Z and Lee K T 2009 Life cycle Stratton R W, Wolfe P J and Hileman J I 2011b Impact of aviation assessment of palm biodiesel: Revealing facts and benefits for Non-CO2 combustion effects on the environmental feasibility of sustainability Appl. Energy 86 S189–96 alternative jet fuels Environ. Sci. Technol. 45 10736–43 Yeh S, Jordaan S M, Brandt A R, Turetsky M R, Spatari S and US Department of Energy 2011 U.S. billion-ton update: biomass Keith D W 2010 Land use greenhouse gas emissions from supply for a bioenergy and bioproducts industry R D Perlack, B J Stokes (Leads), ORNL/TM-2011/224 conventional oil production and oil sands Environ. Sci. Technol. US Environmental Protection Agency 2013 Renewable Fuels 44 8766–72 Standard (Washington, DC)www.epa.gov/otaq/fuels/renewablef Zhou A and Thomson E 2009 The development of biofuels in Asia uels/index.htm Appl. Energy 86 S11–20